Compact Subsets.

Stefano Ferri

September 8, 2024

1 Introduction.
These notes are meant to complement, not to replace, the content of the
textbook [1].
In this notes we shall present the basic definitions and properties of
compact spaces. The concepts are presented for subsets of the real line R but
it should be clear from the statements that most of the theorems presented
can be applied almost without changes to subsets of metric spaces or even
to subsets of topological spaces.
Please, be aware that these are notes written as additional material for
a course and that they did not pass any proofreading other than my own.
If you find any mistake or you think that some part of the notes can be im-
proved, please do not hesitate to contact me at [email protected].

2 Compact Subsets of R.
Definition 2.1. Given any subset ∅ ̸= X ⊂ R a collection U = {Uα : α ∈ I}
of subsets of R which are are open in R and such that
[
Uα ⊇ X
α∈I

is called a open covering of X. Notice that here I is any set of indeces and
that I can be of any cardinality.
Definition 2.2. A subset X of R is compact if given any open covering
U = {Uα : α ∈ I} of X there exists a finite sub-covering of U, i.e. there are
elements Uα1 , . . . , UαN ∈ U such that
N
[
Uαk ⊇ X.
k=1

1
2

Remark 2.3. Since any subset of X is open in X if and only if it is the

intersection of a set U open in R and X we can rephrase the definition of
compactness by saying that X is compact if whenever we have a collection
U = {Uα : α ∈ I} of subsets of X open in X such that
[
X= Uα
α∈I

there are finitely many sets Uα1 , . . . , UαN such that

X = U α1 ∪ · · · ∪ U αN .

In other words, unlike being open or closed, being compact does not depend
on the space in which the set is contained.

Remark 2.4. Notice that in the definition of compact subset of R we are

not saying that X can be covered using finitely many open subsets of X
(this is always true since X is itself open in X) but we are requiring that
given any open covering we can always find finitely many element of that
covering which are still a covering.

By passing to complement sets and remembering that a subset A is closed

in X if and only if A∁ is open in X we can formulate an equivalent definition
of compactness which makes use of closed sets and which is often useful. In
order to do this we need first a preliminary definition.

Definition 2.5. Given any set X we say that a collection A = {Aα :

α ∈ I} of subsets of X has the Finite Intersection Property (in short FIP)
if the intersection of any finite number of elements of A has non-empty
intersection.

Theorem 2.6. A subset X of the real line is compact if and only if every
family of closed subsets of X with the FIP has non-empty intersection.

Proof. Assume first X is a subset of R with the property that every family
of closed subsets of X with the FIP has non-empty intersection. We want
to prove that X is compact. To this end we consider[any open cover {Uα :
α ∈ I}. By definition of covering we have that X = Uα , so by passing to
α∈I \
complement sets and using De Morgan’s Laws we get that ∅ = Uα∁ , with
α∈I
the sets Uα∁ being closed subsets of X. Since we assumed that a collection
of closed subsets of X cannot have empty intersection if it has the FIP it
 3

follows that {Uα∁ : α ∈ I} doesTnot have the FIP. This means that there are
α1 , α2 , . . . , αN ∈ I such that N ∁
i=1 Uαi = ∅. By passing to complement sets
[N
again we get that Uαi = X, the requested finite subcovering of {Uα : α ∈
i=1
I}.
Suppose now that X is compact and T let {Cα : α ∈ I} be a collection of
closed subsets of X and suppose that α∈I Cα = ∅. As before we get that
[
X= Cα∁ , with Cα∁ open, i.e. {Cα∁ : α ∈ I} is an open covering of X. Since
α∈I
X is compact there is a finite subcovering {Cα∁ 1 , Cα∁ 2 , . . . , Cα∁ N } of {Cα∁ : α ∈
N
[ N
\
∁
I}. This means that X = Cαi , and so we have that ∅ = Cαi , which
i=1 i=1
means that the original family of closed sets {Cα : α ∈ I} does not have the
FIP.

Example 2.7. The real line R with the usual metric d(x, y) = |y − x| is not
compact.

Proof. Let U := {(−n, n) : n = 1, 2, . . . }. Because of the Archimedean

property of R for every x ∈ R there is n ∈ N such that n > |x| and so
x ∈ (−n, n), so U is an open covering of R. Now given any finite number of
integers n1 , . . . , nN if we denote M := max{n1 , . . . , nN } we have that
n
[
(−nk , nk ) = (−nM , nM ) ⊊ R.
k=1

So U does not admit any finite subcovering of R and R is not compact.

Example 2.8. Any finite subset X = {x1 , . . . , xn } of R is compact.

Proof. Let U = {Uα : α ∈ I} be an open covering of X. Since it is a covering

for each xi ∈ X there exist a corresponding Uαi ∋ xi (clearly it can happen
that αi = αj for i ̸= j). So {Uαi : i = 1, . . . , n} is a finite subcovering of
U.

Example 2.9. The unit interval [0, 1] ⊂ R with the usual metric is a com-
pact metric space.

Proof. Let U be any open covering of [0, 1] and define

A := {x ∈ [0, 1] : [0, x] can be covered with finitely many elements of U}.

4

Clearly 0 ∈ U for some U ∈ U, so 0 can be actually covered with one

element of U and is therefore in A. Also, if x ∈ A and 0 ≤ x′ < x, then
x′ ∈ A because whichever finite subcollection of U which covers [0, x] covers
also [0, x′ ] ⊂ [0, x]. This means that A is actually a subinterval of [0, 1]
containing 0. If x ∈ A then the same subcollection of U which covers [0, x]
will also cover [0, x + ε] for some ε > 0, if x + ε ∈ [0, 1] because all elements
of U are open sets. This means that A is open in [0, 1]. Moreover, if x is
the right endpoint of A, then x ∈ A because x ∈ Ux for some Ux ∈ U and,
arguing as before, (x − ε, x] ⊂ Ux for some ε > 0, so adding U to the finite
subcollection of U which covers [0, x) we obtain a finite subcollection of U
which covers [0, x]. This means that actually A is also a closed subinterval
of [0, 1], hence A = [0, 1].

Remark 2.10. With small modifications the above proof shows that any
closed subinterval of R is compact.
Before passing to the main theorem of these notes we proof a few pre-
liminary results about compact sets.
Theorem 2.11. Compact subset of the real line are closed in R.
Proof. Let F ⊂ R be compact. We show that F ∁ is open in R. Let x ∈ F ∁ ,
for every point p ∈ F there exists disjoint open intervals Ip ∋ x and Jp ∋ p.
Clearly, {Jp : p ∈ F } is an open covering of F . Hence, there exist p1 , . . . , pN
such that
F ⊆ Jp1 ∪ · · · ∪ JpN .
The set
Ux = Ip1 ∩ · · · ∩ IpN
being a finite Rintersection of open sets is itself open and, clearly, contains x.
Moreover, Ux F = ∅, so Ux is a neighbourhood of x completely contained
in F ∁ , which proves that x is an interior point of F ∁ and that F ∁ is open.

Theorem 2.12. Closed subsets of compact sets are themselves compact.

Proof. Let F ⊂ R be compact and let X a subset of F closed in R. For every
open covering U = {Uα : α ∈ I} of X we can add the set R \ X in order to
obtain an open covering of F . Now using the fact that F is compact we can
find α1 , . . . , αN such that
F ⊆ (R \ X) ∪ Uα1 ∪ · · · ∪ UαN .
This implies that {Uαi : i = 1, 2, . . . , N } is a finite subset of U which still
covers X and hence that X is compact.
 5

Theorem 2.13. For a subset X of R TFAE:

(a.) X is compact;

(b.) Every sequence (xn )∞

n=1 in X admits a convergent subsequence;

(c.) X is closed and bounded.

Proof. We start by proving that (a.) implies (b.).

Assume X to be compact and let (xn )n be a sequence in X. Define for
N ∈ N subsetes CN as:

CN := {xN , xN +1 , xN +2 , . . . }.

The CN are closed subsets of X and, since they are nested, i.e.

C1 ⊃ C2 ⊃ C3 ⊃ . . .

they enjoy the FIP,

\hence, since X is compact they have non-empty inter-
section. Let x0 ∈ CN , we show that we can construct a subsequence of
N ∈N
(xn )n which converges to x0 . Since x0 ∈ CN = {xN , xN +1 , . . . } for every
k ∈ N the interval (x0 − k1 , x0 + k1 ) contains points of {xN , xN +1 , . . . } which
is to say that for every k ∈ N there are infinitely many points of the sequence
(xn )n in (x0 − k1 , x0 + k1 ). To construct the converging subsequence just pick
xn1 ∈ (x0 − 1, x0 + 1), then pick xn2 ∈ (x0 − 12 , x0 + 21 ) with n2 > n1 (you can
do this because the interval (x0 − 12 , x0 + 12 ) contains infinitely many points
of the sequence). Carry on this way by induction: once xn1 , xn2 , . . . , xnk
1 1
have been chosen pick xnk+1 ∈ (x0 − k+1 , x0 + k+1 ) with nk+1 > kk .
The subsequence (xnk )k converges to x0 because given any ε > 0 there
is k0 ∈ N such that k10 < ε and so if k > k0 we have that

1
1


xk ∈ B x0 , ⊆ B x0 , ⊆ B x0 , ε .
k k0

We now prove that (b.) implies (c.).

Suppose now that every sequence in X has a convergent subsequence,
we prove that X must be both closed in R and bounded. If X is not closed
it means that there exists x0 ∈ X \ X and so we can find a sequence (xn )∞
n=1
of points of X which converges to x0 . We know that every subsequence of
a converging sequence converges to the same limit of the original sequence.
6

In this case it means that for every subsequence of (xnk )∞ ∞

k=1 of (xn )n=1 we
have that
lim xnk = x0 ̸∈ X
k→∞

which is against (b.).

If X is not bounded it means that for every n ∈ N there exists xn ∈ X with
|x| > n. We can also find a different element xn for every n, otherwise only
finitely many elements of X would have |x| > n which would imply that
X would be bounded. Every converging subsequence of (xn )∞ n=1 converges
either to ∞ or to −∞ which is again against the assumption (b.).
Finally, we prove that (c.) implies (a.).
Suppose that X is bounded, this means that it is contained in a bounded
closed interval [−M, M ] that, by what we have shown, is compact. Hence X
is a closed subset of R contained in a compact and is therefore compact.

References
[1] Brian S. Thomson, Judith B. Bruckner, Andrew M. Bruck-
ner, Elementary Real Analysis, Prentice Hall (Pearson), 2008.